Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

ILAEDA: An Imitation Learning Based Approach for Automatic Exploratory Data Analysis

Abhijit Manatkar, Devarsh Patel, Hima Patel, Naresh Manwani

TL;DR

This work proposes an AutoEDA model trained through imitation learning from expert EDA sessions, bypassing the need for manually defined interestingness measures, and outperforms the existing state-of-the-art end-to-end EDA approach on benchmarks by up to 3x.

Abstract

Automating end-to-end Exploratory Data Analysis (AutoEDA) is a challenging open problem, often tackled through Reinforcement Learning (RL) by learning to predict a sequence of analysis operations (FILTER, GROUP, etc). Defining rewards for each operation is a challenging task and existing methods rely on various \emph{interestingness measures} to craft reward functions to capture the importance of each operation. In this work, we argue that not all of the essential features of what makes an operation important can be accurately captured mathematically using rewards. We propose an AutoEDA model trained through imitation learning from expert EDA sessions, bypassing the need for manually defined interestingness measures. Our method, based on generative adversarial imitation learning (GAIL), generalizes well across datasets, even with limited expert data. We also introduce a novel approach for generating synthetic EDA demonstrations for training. Our method outperforms the existing state-of-the-art end-to-end EDA approach on benchmarks by upto 3x, showing strong performance and generalization, while naturally capturing diverse interestingness measures in generated EDA sessions.

ILAEDA: An Imitation Learning Based Approach for Automatic Exploratory Data Analysis

TL;DR

This work proposes an AutoEDA model trained through imitation learning from expert EDA sessions, bypassing the need for manually defined interestingness measures, and outperforms the existing state-of-the-art end-to-end EDA approach on benchmarks by up to 3x.

Abstract

Automating end-to-end Exploratory Data Analysis (AutoEDA) is a challenging open problem, often tackled through Reinforcement Learning (RL) by learning to predict a sequence of analysis operations (FILTER, GROUP, etc). Defining rewards for each operation is a challenging task and existing methods rely on various \emph{interestingness measures} to craft reward functions to capture the importance of each operation. In this work, we argue that not all of the essential features of what makes an operation important can be accurately captured mathematically using rewards. We propose an AutoEDA model trained through imitation learning from expert EDA sessions, bypassing the need for manually defined interestingness measures. Our method, based on generative adversarial imitation learning (GAIL), generalizes well across datasets, even with limited expert data. We also introduce a novel approach for generating synthetic EDA demonstrations for training. Our method outperforms the existing state-of-the-art end-to-end EDA approach on benchmarks by upto 3x, showing strong performance and generalization, while naturally capturing diverse interestingness measures in generated EDA sessions.

Paper Structure

This paper contains 32 sections, 8 equations, 2 figures, 8 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. AutoEDA Methods
  4. Imitation Learning
  5. Interestingness Measures
  6. Problem Setup
  7. State Space
  8. Action Space
  9. Issues With Reward-Based Modeling
  10. ILAEDA: Proposed Methodology
  11. Penalties for Incoherent Action Sequences
  12. Initialization of Policy with Behavioral Cloning
  13. Proximal Policy Optimization Updates
  14. Experimental Evaluation
  15. Dataset Details
  16. ...and 17 more sections

Figures (2)

  • Figure 1: ILAEDA System Architecture
  • Figure 2: Distribution of metrics in sessions generated on Synthetic Dataset 7 using ILAEDA models. ILAEDA models were trained on sessions filtered to maximize A-INT, Diversity, or Readability more than other metrics.